PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
IMPACT OF COGNITIVE WORKLOAD ON PHYSIOLOGICAL AROUSAL AND
PERFORMANCE IN YOUNGER AND OLDER DRIVERS
Joonwoo Son1, Bruce Mehler2, Taeyoung Lee1, Yunsuk Park1,
Joseph Coughlin2, & Bryan Reimer2
1
Daegu Gyeongbuk Institute of Science & Technology (DGIST), Daegu, South Korea
2
Massachussets Institute of Technology (MIT) AgeLab & New England University
Transportation Center, Cambridge, Massachusetts, USA
Email: [email protected]
Summary: Two groups, aged 25-35 and 60-69, engaged in 3 levels of a delayed
auditory recall task while driving a simulated highway. Heart rate and skin
conductance increased with each level of demand, demonstrating that these
indices can correctly rank order cognitive workload. Effects were also observed
on speed and SD of lane position, but they were subtle, nonlinear, and did not
effectively differentiate. Patterns were quite consistent across age groups. These
findings on the sensitivity of physiological measures replicate those from an onroad study using a similar protocol. Together, the results support the validity of
using these physiological measures of workload in a simulated environment to
model differences likely to be present under actual driving conditions.
INTRODUCTION
Physiological measures have been proposed as useful metrics for assessing workload in the user
interface design assessment and optimization process (Lenneman & Backs, 2010; Mehler,
Reimer, Coughlin & Dusek, 2009). While physiological indices have been used to measure
workload for some time, particularly in the aviation literature (Wilson, 2002), until recently data
on the sensitivity of such measures in the driving environment has been either limited or
contradictory. An early study (Brookhouis, De Vies & De Waard, 1991) showed that a task
delivered over a cell phone resulted in an increase in heart rate while driving and another
(Brookhouis & De Waard, 2001) showed elevations during heightened driving demands such as
entering a traffic circle. However, an ambitious undertaking involving three different European
centers (HASTE Project) that compared heart rate, heart rate variability and skin conductance
measures for two task types (a surrogate visual interface task and an auditory task) at three levels
of difficulty for each and across a fixed-based simulator, moving-base simulator, and under onroad conditions, produced contradictory results in the different environments and generally either
no significant differentiation across task difficulty levels or counterintuitive response patterns
(Engstrom, Johansson & Ostlund, 2005).
More supportive data have since been presented by Lenneman and Backs (2010) from a
simulation study that monitored a range of cardiovascular measures during single task driving
and dual task conditions using a relatively easy visual n-back target matching task (0-back) and a
much more difficult version (3-back). They found that heart rate was significantly higher during
both the 0-back and the 3-back conditions than during single task driving. The difference in heart
rate between the 0-back and the 3-back was also statistically significant. Cardiac pre-ejection
period (PEP) and a respiratory sinus arrhythmia (RSA) measure were less sensitive, being able to
differentiate single task driving from the 3-back task but not the 0-back task. Another simulation
87
PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
study (Mehler et al., 2009) employed an auditory delayed digit recall task in a variant n-back
format with three levels of task difficulty that allowed a finer grain look at the question of
sensitivity. The research design increased workload in a sequential order from single task driving
through each of the three difficulty levels (0-, 1-, and 2-back) and found that heart rate increased
in a statistically significant manner at each level. Skin conductance level (often used as a
measure of emotional arousal) also increased significantly from single task driving to 0-back,
again from 0-back to the 1-back demand level, and then leveled out during the 2-back. Reimer,
Mehler, Coughlin, Godfrey et al. (2009) used the same auditory recall task and presentation
order in a modest sized on-road evaluation to assess the generalizability of the simulator findings.
They found essentially the same reactivity pattern for heart rate in the field with significant
discrimination between single task driving and 0-back, between 0-back and 1-back, and a further
but non-significant increase during the 2-back. Skin conductance was again sensitive at
indicating initial changes in demand between single task driving and the 0-back but did not
provide further discrimination at the higher levels. Each of these studies used predominately
young adults.
Mehler, Reimer and Coughlin (2010) extended this work by looking at subjects in their 20s, 40s
and 60s to examine whether age impacts the relative sensitivity of heart rate and skin
conductance measures under on-road conditions. They also assessed the extent to which the
sequential presentation of the n-back tasks from easiest to most difficult influenced the apparent
ceiling effect, particularly in skin conductance, at the higher levels of demand. They found that
when the three levels of task difficulty were randomly ordered, and a recovery interval was
provided between tasks, a near linear increase in heart rate appeared across the demand levels.
For both heart rate and skin conductance, each task level was statistically differentiated from
single task driving and from each other. These findings demonstrate that both heart rate and skin
conductance provide the sensitivity to discriminate incremental changes in cognitive workload
induced by this task under actual driving conditions. The study also found that, in relatively
healthy individuals, the pattern of change across demand levels was consistent across age groups.
The present paper examines the extent to which the basic findings observed in Mehler et al.
(2010) are replicable by another research team in a different setting and culture (Korea vs. USA),
with somewhat different age groupings, and in a simulator. Sensitivity in a simulation
environment and the question of the extent to which patterns observed in simulator model what
happens under actual driving conditions, are critical considerations in assessing the degree to
which physiological measures of cognitive workload can be productively used in simulation for
interface design evaluation.
METHOD
Subjects
Subjects were required to meet the following criteria: age between 25-35 or 60-69, drive on
average more than twice a week, be in self-reported good health and free from major medical
conditions, not take medications for psychiatric disorders, score 25 or greater on the mini mental
status exam to establish reasonable cognitive capacity and situational awareness, and have not
88
PROCE
EEDINGS of the Siixth International Driving
D
Symposiu
um on Human Facttors in Driver Asseessment, Training and Vehicle Desiggn
previouslly participateed in a simu
ulated driving
g study. Thee sample connsisted of 30 males: 15 inn the
25-35 age range (M=
=27.9, SD=3.13) and 15 in the 60-699 range (M= 63.2, SD= 11.74).
Experim
mental setup
p
The expeeriment was conducted in
n the DGIST
T fixed-base d driving sim
mulator, whiich incorporated
STISIM Drive™ soft
ftware and a fixed car cab
b (see Figuree 1). Graphiical updates tto the virtuaal
environm
ment were co
omputed usin
ng STISIM Drive™
D
baseed upon inpuuts recordedd from the OE
EM
accelerattor, brake and steering wheel
w
which were
w all auggmented withh tactile forcce feedback. The
virtual ro
oadway was displayed on
n a 2.5m by 2.5m wall-m
mounted screeen at a resoolution of 10224 x
768. Sensory feedbacck to the driv
ver was also
o provided thhrough auditoory and kineetic channelss.
Distance, speed, steering, throttlee, and brakin
ng inputs weere captured at a nominal sampling rrate
of 30 Hz. Physiological data werre collected using
u
a MED
DAC System
m/3 unit and NeuGraph™
™
software (NeuroDynee Medical Corp.,
C
Cambrridge, MA). A display w
was installed on the screeen
he rear-view mirror to pro
ovide inform
mation aboutt the elapsedd time and thhe distance
beside th
remainin
ng in the driv
ve. The simu
ulator configu
uration, STI SIM Drive™
™ protocols, and
physiolog
gical monito
oring equipm
ment and sensor attachmeent methods were all sellected and
implemen
nted in consultation with
h the MIT AgeLab
A
and N
New Englannd Univeristyy Transportaation
Center to
o specifically
y support cro
oss laboratorry, country aand cultural ccollaborativee and replicaation
studies (S
Son, Reimerr, Mehler, Po
ohlmeyer, et al., 2010).
Fiigure 1. The DGIST
D
drivingg simulator
ary task
Seconda
ory delayed digit recall task
t
was useed to create pperiods of coognitive dem
mand at threee
An audito
distinct leevels. This form
f
of n-baack task requ
uires particippants to say oout loud the nth stimulus
back in a sequence th
hat is presentted via audio
o recording. (See Mehleer et al. (20100) for additioonal
backgrou
und and inforrmation on different
d
n-b
back task form
rmats.) The llowest level n-back task is
the 0-bacck where thee participant is to immed
diately repeatt out loud thhe last item ppresented. Att the
moderatee level (1-back), the nextt-to-last stim
muli is to be rrepeated. Att the most difficult level (2back), the second-to--the-last stim
mulus is to bee repeated. T
The n-back w
was administtered as a seeries
of 30 seccond trials co
onsisting of 10 single dig
git numbers (0-9) presennted in a randdomized ordder at
an inter-sstimulus inteerval of 2.1 seconds.
s
Eacch task periood consisted of a set of fo
four trials at a
defined level of difficculty resultin
ng in deman
nd periods thhat were eachh two minutees long.
89
PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
Procedure
Following informed consent, sensor attachment and completion of a pre-experimental
questionnaire, participants received 10 minutes of driving experience and adaptation time in the
simulator. The simulation was then stopped and participants were trained in the n-back task
while remaining seated in the vehicle. N-back training continued until participants met minimum
performance criteria. Performance on the n-back was subsequently formally assessed at each of
the three demand levels with 2 minute breaks between each level. When the simulation was
resumed, participants drove in good weather through 37km of straight highway. Minutes 5
through 7 were used as a single task driving reference (baseline). Thirty seconds later, 18
seconds of instructions introduced the task (0, 1 or 2-back). Each n-back period was 2 minutes in
duration (four 30 second trials). Two minute rest/recovery periods were provided before
presenting instructions for the next task. Presentation order of the three levels of task difficulty
was randomized across participants. A 2 minute interval starting 30 seconds after the last task
was used as the post-task reference period (recovery).
Dependent variables
Error rates on the n-back were used to confirm the extent to which different conditions
represented periods of higher cognitive workload. Error rate was calculated as the percentage of
items that subjects responded to with an incorrect number or give no answer. Average forward
velocity was selected as an indicator of compensatory behavior, since drivers have been observed
to reduce their speed to manage increasing workload (Harms, 1991; Horberry et al., 2006; Son et
al., 2010). Standard deviation of lateral position is another frequently used driving performance
measure (Sayer, Devonshire & Flannagan, 2007). Heart rate and skin conductance level have
been shown to be useful for quantifying changes in workload prior to the occurrence of driving
performance degradation (Lenneman & Backs, 2009; Mehler et al. 2009; 2010).
Analysis method
The significance of age and workload on the dependent variables was assessed. Statistical
comparisons were computed using a repeated measures general linear model (SPSS, ver. 17). A
Greenhouse-Geisser correction was applied for models that violated the assumption of sphericity.
Differences among significant main effects were assessed using pairwise t-tests with a least
significant difference (LSD) adjustment for multiple comparisons.
RESULTS
Secondary Task Performance
Error rates on the 0-back, 1-back and 2-back tasks during the non-driving and driving conditions
appear in Table 1. The overal higher error rates while driving are in line with the position that the
demands of the primary driving task reduced the cognitive resources available to invest in the nback. Under both non-driving conditions and in the dual-task conditon, error rates increased as
the level of cognitive task difficulty increased. Considering the dual task period there was a main
effect of difficulty level on errors (F(1.4,39.3)=55.9, p<.001), a main effect of age (F(1,28)=30.8,
p<.001) reflecting the higher error rate in the older drivers, and a task difficulty by age
90
PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
interaction (F(1.4,39.3)=30.0, p<.001) reflecting the dramatic rise in error rate in the older
drivers relative to the younger group at the 2-back level.
Table 1. Secondary Task Error Scores
Non-driving
Dual Task
Age
0-back
1-back
2-back
0-back
1-back
2-back
25-35
0% (0)
0.74% (2.87)
6.25% (8.35)
0% (0)
4.63% (7.47)
7.08% (6.83)
60-69
0% (0)
3% (4.86)
33.13% (20.86)
0.67% (1.14)
6.48% (6.70) 37.29% (17.46)
* Note: Means with standard deviations in parantheses
Physiological response
Means and standard deviations for heart rate and skin conductance level are listed in Table 2 and
means are displayed graphically in Figure 3. Considering the sample as a whole, there was a
significant effect of demand level on heart rate and skin conductance (F(2.2,62.6)=26.6, p<.001,
F(3.3,91.7)=13.9, p<.001). Compared to the driving baseline period, heart rate during the 0-back
task was 1.5 beats per minute (bpm) higher, 3.4 bpm higher during the 1-back, and 6.5 bpm
higher during the 2-back. Post hoc comparisons show that the incrmental differences between
each of these periods (baseline to 0-back, 0-back to 1-back, 1-back to 2-back) are statistically
distinguishable (p=.002, p<.001, and p=.001 respectively). Skin conductance values during tasks
compared to baseline were 0.5, 0.9, and 1.1 micromhos higher for the 0-back, 1-back, and 2-back
periods respectively. Post hoc comparisons show that the incremental differences between
baseline and 0-back and 0-back and 1-back are statistically distinguishable (p=.015 and p=.007).
Mean skin conductance for the 2-back was higher than during the 1-back, but this difference was
not statistically significant (p=.140). While means were consistently lower for the older drivers,
and the absolute magnitude of change was somewhat less, there was no statistically significant
effect of age on heart rate and skin conductance (F(1,28)=1.4, p=.251 and F(1,28)=0.5, p=.467).
Table 2. Summary of Physiological Response Measures During Simulated Driving
Heart Rate (beats/min)
Younger (25-35)
Skin Conductance Level (micromhos)
Older (60-69)
Younger (25-35)
Older (60-69)
Baseline
78.5 (15.4)
74.6 (11.1)
7.5 (3.5)
6.9 (3.1)
0-back
80.9 (15.1)
75.2 (10.1)
8.1 (3.8)
7.2 (3.2)
1-back
83.2 (15.1)
76.7 (10.1)
8.7 (3.9)
7.5 (3.2)
2-back
86.3 (15.9)
79.8 (11.8)
8.9 (3.9)
7.6 (3.4)
Recovery
77.9 (14.4)
73.2 ( 9.1)
7.9 (3.9)
7.1 (2.8)
* Note: mean (standard deviation)
91
PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
10
25-35
Average Skin Conductance Level
(micromhos/min)
Average Heart Rate (beats/min)
90
60-69
85
80
75
70
25-35
60-69
9
8
7
6
Baseline
0-back
1-back
2-back
Recovery
Baseline
0-back
1-back
2-back
Recovery
Figure 3. Mean Physiological Arousal as a Function of Task Level:
Heart Rate (left) and Skin Conductance Level (right)
Vehicle Control
To observe the compensatory behaviors and perfomance changes under different levels of
cognitive workload by age, forward velocity and standard deviation of lane position were
examined. As shown in Figure 4, both age groups decreased vehicle speed under the dual task
conditions. Although the average velocity and the standard deviation of lane position profiles did
not show a simple correlation with the level of cognitive workload, the driving performance was
impacted during the dual task conditions as show in an overall effect of period (velocity: F(3.1,
85.9)=13.5, p<.001; SDLP: F(3.4, 96.1)=6.3, p<.001).
0.08
25-35
25-35
60-69
60-69
S.D. of Lane Position (m)
Average Velocity (kph)
106
102
98
94
0.06
0.04
0.02
90
Baseline
0-back
1-back
2-back
Baseline
Recovery
0-back
1-back
2-back
Recovery
Figure 4. Driving Performance Measures as a Function of Task Level:
Average Velocity (left) and Standard Deviation (SD) of Lane Position (right)
DISCUSSION
N-back errors rates from both the non-driving period and while driving the simulator clearly
document an incremental increase in difficulty across the task levels. Coincident with this, heart
rate and skin conductance during simulated driving show an unambiguous increase in mean
value for each level of heightened demand. These findings closely parallel what has been
observed under field conditions (Mehler et al., 2010) and further strengthen the position that
simulation can be used to validly model physiological reactivity patterns in response to
incremental differences in cognitive workload (Reimer & Mehler, 2010). In addition, this study
extends the available data on age and physiological activation while driving. As was the case in
Mehler et al. (2010), individual variability in heart rate and skin conductance level are large
enough (see standard deviation values in Table 2) that chronological age is not a statistically
92
PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
significant factor in relatively healthy individuals up through the age of 69 who have been selfreport screened for major medical complications that might affect reactivity. It remains to be
determined to what extent the screened medications and health conditions would impact this
finding. The pattern of change in heart rate and skin conductance level with increasing mental
workload differs in important ways from what is seen in the two driving performance measures.
Compensatory slowing of driving speed is observed during the dual task periods, yet the
relationship to level of demand is not linear and could not be used, for example, to discriminate
between the 0-back and 2-back tasks. Similarly, the impact of cognitive workload on standard
deviation of lane position could not be used to correctly rank order task demand. These results
support the arguments by Lenneman and Backs (2010) and Mehler et al. (2009) that selected
physiological measures are more useful for detecting relative changes in mental workload at
levels below which clear decrements in driving performance begin to be observed.
ACKNOWLEDGMENTS
This research was funded primarily through the DGIST Research Program of the Korea Ministry
of Education, Science and Technology (MEST), with additional support from the USDOT’s
Region I New England University Transportation Center.
REFERENCES
Brookhuis, K. A. & De Waard, D. (2001). Assessment of drivers' workload: Performance and
subjective and physiological indexes. In P. A. Hancock & P. A. Desmond (Eds.), Stress,
Workload, and Fatigue (pp. 321-333). Mahwah, NJ: Lawrence Erlbaum Associates.
Brookhuis, K. A., de Vries, G. & de Waard, D. (1991). The effects of mobile telephoning on
driving performance. Accident Analysis and Prevention, 23, 309-316.
Engström, J., Johansson, E., & Östlund, J. (2005). Effects of visual and cognitive load in real and
simulated motorway driving. Transportation Research Part F, 8(2), 97-120.
Harms, L. (1991). Variation in drivers' cognitive load: Effects of driving through village areas
and rural junctions. Ergonomics, 34(2), 151-160.
Horberry, T., Anderson, J., Regan, M. A., Triggs, T.J., & Brown, J. (2006). Driver distraction:
The effects of concurrent in-vehicle tasks, road environment complexity and age on driving
performance. Accident Analysis and Prevention, 38(1), 185-191.
Lenneman, J.K. & Backs, R.W. (2009). Cardiac autonomic control during simulated driving with
a concurrent verbal working memory task. Human Factors, 53(3), 404-418.
Lenneman, J.K. & Backs, R.W. (2010). Enhancing assessment of in-vehicle technology attention
demands with cardiac measures. Proceedings of the Second International Conference on
Automotive User Interfaces and Interactive Vehicular Applications (AutoUI 2010),
November 11-12, Pittsburgh, Pennsylvania, USA.
Mehler, B., Reimer, B., Coughlin, J.F., & Dusek, J.A. (2009). The impact of incremental
increases in cognitive workload on physiological arousal and performance in young adult
drivers. Transportation Research Record, 2138, 6-12.
Mehler, B., Reimer, B., & Coughlin, J.F. (2010). Physiological reactivity to graded levels of
cognitive workload across three age groups: An on-road evaluation. Proceedings of the
93
PROCEEDINGS of the Sixth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design
Human Factors and Ergonomics Society 54th Annual Meeting, San Francisco, CA, 20622066.
Reimer, B. & Mehler, B. (2010). The impact of cognitive workload on physiological arousal in
young adult drivers: a field study and simulation validation. Manuscript under review.
Reimer, B., Mehler, B., Coughlin, J. F., Godfrey, K. M., & Tan, C. (2009). An on-road
assessment of the impact of cognitive workload on physiological arousal in young adult
drivers. Proceedings of the AutomotiveUI, Essen, Germany.
Reimer, B., Mehler, B., Wang, Y., & Coughlin, J.F. (2010). The impact of systematic variation
of cognitive demand on drivers’ visual attention across multiple age groups. Proceedings of
the 54th Annual Meeting of the Human Factors and Ergonomics Society, San Francisco, Sept.
27-Oct. 1, 2010, 2052-2056.
Sayer, J., Devonshire, J. & Flannagan, C. (2007). Naturalistic driving performance during
secondary tasks. Proceedings of the Fourth International Driving Symposium on Human
Factors in Driver Assessment, Training and Vehicle Design, 224-230.
Son, J., Reimer, B., Mehler, B., Pohlmeyer, A. E., Godfrey, K.M., Orszulak, J., Long, J., Kim,
M.H., Lee, Y.T. & Coughlin, J.F. (2010). Age and cross-cultural comparison of drivers’
cognitive workload and performance in simulated urban driving. International Journal of
Automotive Technology, 11(4), 533-539.
Wilson, G. F. (2002). Psychophysiological test methods and procedures. In S. G. Charlton & T.
G. O’Brien (Eds.), Handbook of Human Factors Testing and Evaluation (pp. 127-156).
Mahwah, NJ: Lawrence Erlbaum.
94